基于自适应神经网络的有源电力滤波器谐波电流提取方法

Harmonic detection based on adaptive neural network for active power filter

对有源电力滤波器(APF)谐波检测方法进行研究,将自适应神经网络(ADNN)的信号处理技术应用于三相非线性负载的谐波电流检测中,提出了一种基于ADNN检测方法。该方法的主要特点是不仅能检测谐波电流中的基波成分,也能检测出其中不同的高次谐波分量的幅值和相位。通过搭建一个基于DSP三相并联APF的实验系统来证实该方法的有效性,并与传统的傅里叶变换法和d-q算法相比较,验证其具有更快的动态响应。最后,通过分别对感性负载和容性负载进行阶跃变化操作试验检验所提出的ADNN检测法的APF谐波补偿特性,并获得了优良的补偿性能。

The harmonic current detection methods of APF （Active Power Filter） are studied and a ADNN-based detection method is proposed,which applies ADNN signal-processing technique to detect the total harmonic current of three-phase nonlinear load,not only its fundamental component but also the amplitude and phase of its high-order harmonic components. The validity of proposed method is verified with a DSP-based experimental three-phase shunt APF and its faster dynamic response is proved by comparing with traditional methods of Fourier transform and d-q algorithm. The better compensation performance of ADNN-based APF is verified by the step change tests respectively with inductive load and capacitive load.